net/sunrpc/xprt.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/net/sunrpc/xprt.c
   4  *
   5  *  This is a generic RPC call interface supporting congestion avoidance,
   6  *  and asynchronous calls.
   7  *
   8  *  The interface works like this:
   9  *
  10  *  -   When a process places a call, it allocates a request slot if
  11  *      one is available. Otherwise, it sleeps on the backlog queue
  12  *      (xprt_reserve).
  13  *  -   Next, the caller puts together the RPC message, stuffs it into
  14  *      the request struct, and calls xprt_transmit().
  15  *  -   xprt_transmit sends the message and installs the caller on the
  16  *      transport's wait list. At the same time, if a reply is expected,
  17  *      it installs a timer that is run after the packet's timeout has
  18  *      expired.
  19  *  -   When a packet arrives, the data_ready handler walks the list of
  20  *      pending requests for that transport. If a matching XID is found, the
  21  *      caller is woken up, and the timer removed.
  22  *  -   When no reply arrives within the timeout interval, the timer is
  23  *      fired by the kernel and runs xprt_timer(). It either adjusts the
  24  *      timeout values (minor timeout) or wakes up the caller with a status
  25  *      of -ETIMEDOUT.
  26  *  -   When the caller receives a notification from RPC that a reply arrived,
  27  *      it should release the RPC slot, and process the reply.
  28  *      If the call timed out, it may choose to retry the operation by
  29  *      adjusting the initial timeout value, and simply calling rpc_call
  30  *      again.
  31  *
  32  *  Support for async RPC is done through a set of RPC-specific scheduling
  33  *  primitives that `transparently' work for processes as well as async
  34  *  tasks that rely on callbacks.
  35  *
  36  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
  37  *
  38  *  Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
  39  */
  40
  41 #include <linux/module.h>
  42
  43 #include <linux/types.h>
  44 #include <linux/interrupt.h>
  45 #include <linux/workqueue.h>
  46 #include <linux/net.h>
  47 #include <linux/ktime.h>
  48
  49 #include <linux/sunrpc/clnt.h>
  50 #include <linux/sunrpc/metrics.h>
  51 #include <linux/sunrpc/bc_xprt.h>
  52 #include <linux/rcupdate.h>
  53 #include <linux/sched/mm.h>
  54
  55 #include <trace/events/sunrpc.h>
  56
  57 #include "sunrpc.h"
  58
  59 /*
  60  * Local variables
  61  */
  62
  63 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
  64 # define RPCDBG_FACILITY        RPCDBG_XPRT
  65 #endif
  66
  67 /*
  68  * Local functions
  69  */
  70 static void      xprt_init(struct rpc_xprt *xprt, struct net *net);
  71 static __be32   xprt_alloc_xid(struct rpc_xprt *xprt);
  72 static void      xprt_destroy(struct rpc_xprt *xprt);
  73
  74 static DEFINE_SPINLOCK(xprt_list_lock);
  75 static LIST_HEAD(xprt_list);
  76
  77 static unsigned long xprt_request_timeout(const struct rpc_rqst *req)
  78 {
  79         unsigned long timeout = jiffies + req->rq_timeout;
  80
  81         if (time_before(timeout, req->rq_majortimeo))
  82                 return timeout;
  83         return req->rq_majortimeo;
  84 }
  85
  86 /**
  87  * xprt_register_transport - register a transport implementation
  88  * @transport: transport to register
  89  *
  90  * If a transport implementation is loaded as a kernel module, it can
  91  * call this interface to make itself known to the RPC client.
  92  *
  93  * Returns:
  94  * 0:           transport successfully registered
  95  * -EEXIST:     transport already registered
  96  * -EINVAL:     transport module being unloaded
  97  */
  98 int xprt_register_transport(struct xprt_class *transport)
  99 {
 100         struct xprt_class *t;
 101         int result;
 102
 103         result = -EEXIST;
 104         spin_lock(&xprt_list_lock);
 105         list_for_each_entry(t, &xprt_list, list) {
 106                 /* don't register the same transport class twice */
 107                 if (t->ident == transport->ident)
 108                         goto out;
 109         }
 110
 111         list_add_tail(&transport->list, &xprt_list);
 112         printk(KERN_INFO "RPC: Registered %s transport module.\n",
 113                transport->name);
 114         result = 0;
 115
 116 out:
 117         spin_unlock(&xprt_list_lock);
 118         return result;
 119 }
 120 EXPORT_SYMBOL_GPL(xprt_register_transport);
 121
 122 /**
 123  * xprt_unregister_transport - unregister a transport implementation
 124  * @transport: transport to unregister
 125  *
 126  * Returns:
 127  * 0:           transport successfully unregistered
 128  * -ENOENT:     transport never registered
 129  */
 130 int xprt_unregister_transport(struct xprt_class *transport)
 131 {
 132         struct xprt_class *t;
 133         int result;
 134
 135         result = 0;
 136         spin_lock(&xprt_list_lock);
 137         list_for_each_entry(t, &xprt_list, list) {
 138                 if (t == transport) {
 139                         printk(KERN_INFO
 140                                 "RPC: Unregistered %s transport module.\n",
 141                                 transport->name);
 142                         list_del_init(&transport->list);
 143                         goto out;
 144                 }
 145         }
 146         result = -ENOENT;
 147
 148 out:
 149         spin_unlock(&xprt_list_lock);
 150         return result;
 151 }
 152 EXPORT_SYMBOL_GPL(xprt_unregister_transport);
 153
 154 /**
 155  * xprt_load_transport - load a transport implementation
 156  * @transport_name: transport to load
 157  *
 158  * Returns:
 159  * 0:           transport successfully loaded
 160  * -ENOENT:     transport module not available
 161  */
 162 int xprt_load_transport(const char *transport_name)
 163 {
 164         struct xprt_class *t;
 165         int result;
 166
 167         result = 0;
 168         spin_lock(&xprt_list_lock);
 169         list_for_each_entry(t, &xprt_list, list) {
 170                 if (strcmp(t->name, transport_name) == 0) {
 171                         spin_unlock(&xprt_list_lock);
 172                         goto out;
 173                 }
 174         }
 175         spin_unlock(&xprt_list_lock);
 176         result = request_module("xprt%s", transport_name);
 177 out:
 178         return result;
 179 }
 180 EXPORT_SYMBOL_GPL(xprt_load_transport);
 181
 182 static void xprt_clear_locked(struct rpc_xprt *xprt)
 183 {
 184         xprt->snd_task = NULL;
 185         if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state)) {
 186                 smp_mb__before_atomic();
 187                 clear_bit(XPRT_LOCKED, &xprt->state);
 188                 smp_mb__after_atomic();
 189         } else
 190                 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 191 }
 192
 193 /**
 194  * xprt_reserve_xprt - serialize write access to transports
 195  * @task: task that is requesting access to the transport
 196  * @xprt: pointer to the target transport
 197  *
 198  * This prevents mixing the payload of separate requests, and prevents
 199  * transport connects from colliding with writes.  No congestion control
 200  * is provided.
 201  */
 202 int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 203 {
 204         struct rpc_rqst *req = task->tk_rqstp;
 205
 206         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
 207                 if (task == xprt->snd_task)
 208                         return 1;
 209                 goto out_sleep;
 210         }
 211         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 212                 goto out_unlock;
 213         xprt->snd_task = task;
 214
 215         return 1;
 216
 217 out_unlock:
 218         xprt_clear_locked(xprt);
 219 out_sleep:
 220         dprintk("RPC: %5u failed to lock transport %p\n",
 221                         task->tk_pid, xprt);
 222         task->tk_status = -EAGAIN;
 223         if  (RPC_IS_SOFT(task))
 224                 rpc_sleep_on_timeout(&xprt->sending, task, NULL,
 225                                 xprt_request_timeout(req));
 226         else
 227                 rpc_sleep_on(&xprt->sending, task, NULL);
 228         return 0;
 229 }
 230 EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
 231
 232 static bool
 233 xprt_need_congestion_window_wait(struct rpc_xprt *xprt)
 234 {
 235         return test_bit(XPRT_CWND_WAIT, &xprt->state);
 236 }
 237
 238 static void
 239 xprt_set_congestion_window_wait(struct rpc_xprt *xprt)
 240 {
 241         if (!list_empty(&xprt->xmit_queue)) {
 242                 /* Peek at head of queue to see if it can make progress */
 243                 if (list_first_entry(&xprt->xmit_queue, struct rpc_rqst,
 244                                         rq_xmit)->rq_cong)
 245                         return;
 246         }
 247         set_bit(XPRT_CWND_WAIT, &xprt->state);
 248 }
 249
 250 static void
 251 xprt_test_and_clear_congestion_window_wait(struct rpc_xprt *xprt)
 252 {
 253         if (!RPCXPRT_CONGESTED(xprt))
 254                 clear_bit(XPRT_CWND_WAIT, &xprt->state);
 255 }
 256
 257 /*
 258  * xprt_reserve_xprt_cong - serialize write access to transports
 259  * @task: task that is requesting access to the transport
 260  *
 261  * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
 262  * integrated into the decision of whether a request is allowed to be
 263  * woken up and given access to the transport.
 264  * Note that the lock is only granted if we know there are free slots.
 265  */
 266 int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 267 {
 268         struct rpc_rqst *req = task->tk_rqstp;
 269
 270         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
 271                 if (task == xprt->snd_task)
 272                         return 1;
 273                 goto out_sleep;
 274         }
 275         if (req == NULL) {
 276                 xprt->snd_task = task;
 277                 return 1;
 278         }
 279         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 280                 goto out_unlock;
 281         if (!xprt_need_congestion_window_wait(xprt)) {
 282                 xprt->snd_task = task;
 283                 return 1;
 284         }
 285 out_unlock:
 286         xprt_clear_locked(xprt);
 287 out_sleep:
 288         dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
 289         task->tk_status = -EAGAIN;
 290         if (RPC_IS_SOFT(task))
 291                 rpc_sleep_on_timeout(&xprt->sending, task, NULL,
 292                                 xprt_request_timeout(req));
 293         else
 294                 rpc_sleep_on(&xprt->sending, task, NULL);
 295         return 0;
 296 }
 297 EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
 298
 299 static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
 300 {
 301         int retval;
 302
 303         if (test_bit(XPRT_LOCKED, &xprt->state) && xprt->snd_task == task)
 304                 return 1;
 305         spin_lock(&xprt->transport_lock);
 306         retval = xprt->ops->reserve_xprt(xprt, task);
 307         spin_unlock(&xprt->transport_lock);
 308         return retval;
 309 }
 310
 311 static bool __xprt_lock_write_func(struct rpc_task *task, void *data)
 312 {
 313         struct rpc_xprt *xprt = data;
 314
 315         xprt->snd_task = task;
 316         return true;
 317 }
 318
 319 static void __xprt_lock_write_next(struct rpc_xprt *xprt)
 320 {
 321         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 322                 return;
 323         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 324                 goto out_unlock;
 325         if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
 326                                 __xprt_lock_write_func, xprt))
 327                 return;
 328 out_unlock:
 329         xprt_clear_locked(xprt);
 330 }
 331
 332 static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
 333 {
 334         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 335                 return;
 336         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 337                 goto out_unlock;
 338         if (xprt_need_congestion_window_wait(xprt))
 339                 goto out_unlock;
 340         if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
 341                                 __xprt_lock_write_func, xprt))
 342                 return;
 343 out_unlock:
 344         xprt_clear_locked(xprt);
 345 }
 346
 347 /**
 348  * xprt_release_xprt - allow other requests to use a transport
 349  * @xprt: transport with other tasks potentially waiting
 350  * @task: task that is releasing access to the transport
 351  *
 352  * Note that "task" can be NULL.  No congestion control is provided.
 353  */
 354 void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 355 {
 356         if (xprt->snd_task == task) {
 357                 xprt_clear_locked(xprt);
 358                 __xprt_lock_write_next(xprt);
 359         }
 360 }
 361 EXPORT_SYMBOL_GPL(xprt_release_xprt);
 362
 363 /**
 364  * xprt_release_xprt_cong - allow other requests to use a transport
 365  * @xprt: transport with other tasks potentially waiting
 366  * @task: task that is releasing access to the transport
 367  *
 368  * Note that "task" can be NULL.  Another task is awoken to use the
 369  * transport if the transport's congestion window allows it.
 370  */
 371 void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 372 {
 373         if (xprt->snd_task == task) {
 374                 xprt_clear_locked(xprt);
 375                 __xprt_lock_write_next_cong(xprt);
 376         }
 377 }
 378 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
 379
 380 static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 381 {
 382         if (xprt->snd_task != task)
 383                 return;
 384         spin_lock(&xprt->transport_lock);
 385         xprt->ops->release_xprt(xprt, task);
 386         spin_unlock(&xprt->transport_lock);
 387 }
 388
 389 /*
 390  * Van Jacobson congestion avoidance. Check if the congestion window
 391  * overflowed. Put the task to sleep if this is the case.
 392  */
 393 static int
 394 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 395 {
 396         if (req->rq_cong)
 397                 return 1;
 398         dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
 399                         req->rq_task->tk_pid, xprt->cong, xprt->cwnd);
 400         if (RPCXPRT_CONGESTED(xprt)) {
 401                 xprt_set_congestion_window_wait(xprt);
 402                 return 0;
 403         }
 404         req->rq_cong = 1;
 405         xprt->cong += RPC_CWNDSCALE;
 406         return 1;
 407 }
 408
 409 /*
 410  * Adjust the congestion window, and wake up the next task
 411  * that has been sleeping due to congestion
 412  */
 413 static void
 414 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 415 {
 416         if (!req->rq_cong)
 417                 return;
 418         req->rq_cong = 0;
 419         xprt->cong -= RPC_CWNDSCALE;
 420         xprt_test_and_clear_congestion_window_wait(xprt);
 421         __xprt_lock_write_next_cong(xprt);
 422 }
 423
 424 /**
 425  * xprt_request_get_cong - Request congestion control credits
 426  * @xprt: pointer to transport
 427  * @req: pointer to RPC request
 428  *
 429  * Useful for transports that require congestion control.
 430  */
 431 bool
 432 xprt_request_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 433 {
 434         bool ret = false;
 435
 436         if (req->rq_cong)
 437                 return true;
 438         spin_lock(&xprt->transport_lock);
 439         ret = __xprt_get_cong(xprt, req) != 0;
 440         spin_unlock(&xprt->transport_lock);
 441         return ret;
 442 }
 443 EXPORT_SYMBOL_GPL(xprt_request_get_cong);
 444
 445 /**
 446  * xprt_release_rqst_cong - housekeeping when request is complete
 447  * @task: RPC request that recently completed
 448  *
 449  * Useful for transports that require congestion control.
 450  */
 451 void xprt_release_rqst_cong(struct rpc_task *task)
 452 {
 453         struct rpc_rqst *req = task->tk_rqstp;
 454
 455         __xprt_put_cong(req->rq_xprt, req);
 456 }
 457 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
 458
 459 /*
 460  * Clear the congestion window wait flag and wake up the next
 461  * entry on xprt->sending
 462  */
 463 static void
 464 xprt_clear_congestion_window_wait(struct rpc_xprt *xprt)
 465 {
 466         if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state)) {
 467                 spin_lock(&xprt->transport_lock);
 468                 __xprt_lock_write_next_cong(xprt);
 469                 spin_unlock(&xprt->transport_lock);
 470         }
 471 }
 472
 473 /**
 474  * xprt_adjust_cwnd - adjust transport congestion window
 475  * @xprt: pointer to xprt
 476  * @task: recently completed RPC request used to adjust window
 477  * @result: result code of completed RPC request
 478  *
 479  * The transport code maintains an estimate on the maximum number of out-
 480  * standing RPC requests, using a smoothed version of the congestion
 481  * avoidance implemented in 44BSD. This is basically the Van Jacobson
 482  * congestion algorithm: If a retransmit occurs, the congestion window is
 483  * halved; otherwise, it is incremented by 1/cwnd when
 484  *
 485  *      -       a reply is received and
 486  *      -       a full number of requests are outstanding and
 487  *      -       the congestion window hasn't been updated recently.
 488  */
 489 void xprt_adjust_cwnd(struct rpc_xprt *xprt, struct rpc_task *task, int result)
 490 {
 491         struct rpc_rqst *req = task->tk_rqstp;
 492         unsigned long cwnd = xprt->cwnd;
 493
 494         if (result >= 0 && cwnd <= xprt->cong) {
 495                 /* The (cwnd >> 1) term makes sure
 496                  * the result gets rounded properly. */
 497                 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
 498                 if (cwnd > RPC_MAXCWND(xprt))
 499                         cwnd = RPC_MAXCWND(xprt);
 500                 __xprt_lock_write_next_cong(xprt);
 501         } else if (result == -ETIMEDOUT) {
 502                 cwnd >>= 1;
 503                 if (cwnd < RPC_CWNDSCALE)
 504                         cwnd = RPC_CWNDSCALE;
 505         }
 506         dprintk("RPC:       cong %ld, cwnd was %ld, now %ld\n",
 507                         xprt->cong, xprt->cwnd, cwnd);
 508         xprt->cwnd = cwnd;
 509         __xprt_put_cong(xprt, req);
 510 }
 511 EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
 512
 513 /**
 514  * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
 515  * @xprt: transport with waiting tasks
 516  * @status: result code to plant in each task before waking it
 517  *
 518  */
 519 void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
 520 {
 521         if (status < 0)
 522                 rpc_wake_up_status(&xprt->pending, status);
 523         else
 524                 rpc_wake_up(&xprt->pending);
 525 }
 526 EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
 527
 528 /**
 529  * xprt_wait_for_buffer_space - wait for transport output buffer to clear
 530  * @xprt: transport
 531  *
 532  * Note that we only set the timer for the case of RPC_IS_SOFT(), since
 533  * we don't in general want to force a socket disconnection due to
 534  * an incomplete RPC call transmission.
 535  */
 536 void xprt_wait_for_buffer_space(struct rpc_xprt *xprt)
 537 {
 538         set_bit(XPRT_WRITE_SPACE, &xprt->state);
 539 }
 540 EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
 541
 542 static bool
 543 xprt_clear_write_space_locked(struct rpc_xprt *xprt)
 544 {
 545         if (test_and_clear_bit(XPRT_WRITE_SPACE, &xprt->state)) {
 546                 __xprt_lock_write_next(xprt);
 547                 dprintk("RPC:       write space: waking waiting task on "
 548                                 "xprt %p\n", xprt);
 549                 return true;
 550         }
 551         return false;
 552 }
 553
 554 /**
 555  * xprt_write_space - wake the task waiting for transport output buffer space
 556  * @xprt: transport with waiting tasks
 557  *
 558  * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
 559  */
 560 bool xprt_write_space(struct rpc_xprt *xprt)
 561 {
 562         bool ret;
 563
 564         if (!test_bit(XPRT_WRITE_SPACE, &xprt->state))
 565                 return false;
 566         spin_lock(&xprt->transport_lock);
 567         ret = xprt_clear_write_space_locked(xprt);
 568         spin_unlock(&xprt->transport_lock);
 569         return ret;
 570 }
 571 EXPORT_SYMBOL_GPL(xprt_write_space);
 572
 573 static unsigned long xprt_abs_ktime_to_jiffies(ktime_t abstime)
 574 {
 575         s64 delta = ktime_to_ns(ktime_get() - abstime);
 576         return likely(delta >= 0) ?
 577                 jiffies - nsecs_to_jiffies(delta) :
 578                 jiffies + nsecs_to_jiffies(-delta);
 579 }
 580
 581 static unsigned long xprt_calc_majortimeo(struct rpc_rqst *req)
 582 {
 583         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
 584         unsigned long majortimeo = req->rq_timeout;
 585
 586         if (to->to_exponential)
 587                 majortimeo <<= to->to_retries;
 588         else
 589                 majortimeo += to->to_increment * to->to_retries;
 590         if (majortimeo > to->to_maxval || majortimeo == 0)
 591                 majortimeo = to->to_maxval;
 592         return majortimeo;
 593 }
 594
 595 static void xprt_reset_majortimeo(struct rpc_rqst *req)
 596 {
 597         req->rq_majortimeo += xprt_calc_majortimeo(req);
 598 }
 599
 600 static void xprt_init_majortimeo(struct rpc_task *task, struct rpc_rqst *req)
 601 {
 602         unsigned long time_init;
 603         struct rpc_xprt *xprt = req->rq_xprt;
 604
 605         if (likely(xprt && xprt_connected(xprt)))
 606                 time_init = jiffies;
 607         else
 608                 time_init = xprt_abs_ktime_to_jiffies(task->tk_start);
 609         req->rq_timeout = task->tk_client->cl_timeout->to_initval;
 610         req->rq_majortimeo = time_init + xprt_calc_majortimeo(req);
 611 }
 612
 613 /**
 614  * xprt_adjust_timeout - adjust timeout values for next retransmit
 615  * @req: RPC request containing parameters to use for the adjustment
 616  *
 617  */
 618 int xprt_adjust_timeout(struct rpc_rqst *req)
 619 {
 620         struct rpc_xprt *xprt = req->rq_xprt;
 621         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
 622         int status = 0;
 623
 624         if (time_before(jiffies, req->rq_majortimeo)) {
 625                 if (to->to_exponential)
 626                         req->rq_timeout <<= 1;
 627                 else
 628                         req->rq_timeout += to->to_increment;
 629                 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
 630                         req->rq_timeout = to->to_maxval;
 631                 req->rq_retries++;
 632         } else {
 633                 req->rq_timeout = to->to_initval;
 634                 req->rq_retries = 0;
 635                 xprt_reset_majortimeo(req);
 636                 /* Reset the RTT counters == "slow start" */
 637                 spin_lock(&xprt->transport_lock);
 638                 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
 639                 spin_unlock(&xprt->transport_lock);
 640                 status = -ETIMEDOUT;
 641         }
 642
 643         if (req->rq_timeout == 0) {
 644                 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
 645                 req->rq_timeout = 5 * HZ;
 646         }
 647         return status;
 648 }
 649
 650 static void xprt_autoclose(struct work_struct *work)
 651 {
 652         struct rpc_xprt *xprt =
 653                 container_of(work, struct rpc_xprt, task_cleanup);
 654         unsigned int pflags = memalloc_nofs_save();
 655
 656         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
 657         xprt->ops->close(xprt);
 658         xprt_release_write(xprt, NULL);
 659         wake_up_bit(&xprt->state, XPRT_LOCKED);
 660         memalloc_nofs_restore(pflags);
 661 }
 662
 663 /**
 664  * xprt_disconnect_done - mark a transport as disconnected
 665  * @xprt: transport to flag for disconnect
 666  *
 667  */
 668 void xprt_disconnect_done(struct rpc_xprt *xprt)
 669 {
 670         dprintk("RPC:       disconnected transport %p\n", xprt);
 671         spin_lock(&xprt->transport_lock);
 672         xprt_clear_connected(xprt);
 673         xprt_clear_write_space_locked(xprt);
 674         xprt_wake_pending_tasks(xprt, -ENOTCONN);
 675         spin_unlock(&xprt->transport_lock);
 676 }
 677 EXPORT_SYMBOL_GPL(xprt_disconnect_done);
 678
 679 /**
 680  * xprt_force_disconnect - force a transport to disconnect
 681  * @xprt: transport to disconnect
 682  *
 683  */
 684 void xprt_force_disconnect(struct rpc_xprt *xprt)
 685 {
 686         /* Don't race with the test_bit() in xprt_clear_locked() */
 687         spin_lock(&xprt->transport_lock);
 688         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 689         /* Try to schedule an autoclose RPC call */
 690         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 691                 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 692         else if (xprt->snd_task)
 693                 rpc_wake_up_queued_task_set_status(&xprt->pending,
 694                                 xprt->snd_task, -ENOTCONN);
 695         spin_unlock(&xprt->transport_lock);
 696 }
 697 EXPORT_SYMBOL_GPL(xprt_force_disconnect);
 698
 699 static unsigned int
 700 xprt_connect_cookie(struct rpc_xprt *xprt)
 701 {
 702         return READ_ONCE(xprt->connect_cookie);
 703 }
 704
 705 static bool
 706 xprt_request_retransmit_after_disconnect(struct rpc_task *task)
 707 {
 708         struct rpc_rqst *req = task->tk_rqstp;
 709         struct rpc_xprt *xprt = req->rq_xprt;
 710
 711         return req->rq_connect_cookie != xprt_connect_cookie(xprt) ||
 712                 !xprt_connected(xprt);
 713 }
 714
 715 /**
 716  * xprt_conditional_disconnect - force a transport to disconnect
 717  * @xprt: transport to disconnect
 718  * @cookie: 'connection cookie'
 719  *
 720  * This attempts to break the connection if and only if 'cookie' matches
 721  * the current transport 'connection cookie'. It ensures that we don't
 722  * try to break the connection more than once when we need to retransmit
 723  * a batch of RPC requests.
 724  *
 725  */
 726 void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
 727 {
 728         /* Don't race with the test_bit() in xprt_clear_locked() */
 729         spin_lock(&xprt->transport_lock);
 730         if (cookie != xprt->connect_cookie)
 731                 goto out;
 732         if (test_bit(XPRT_CLOSING, &xprt->state))
 733                 goto out;
 734         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 735         /* Try to schedule an autoclose RPC call */
 736         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 737                 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 738         xprt_wake_pending_tasks(xprt, -EAGAIN);
 739 out:
 740         spin_unlock(&xprt->transport_lock);
 741 }
 742
 743 static bool
 744 xprt_has_timer(const struct rpc_xprt *xprt)
 745 {
 746         return xprt->idle_timeout != 0;
 747 }
 748
 749 static void
 750 xprt_schedule_autodisconnect(struct rpc_xprt *xprt)
 751         __must_hold(&xprt->transport_lock)
 752 {
 753         xprt->last_used = jiffies;
 754         if (RB_EMPTY_ROOT(&xprt->recv_queue) && xprt_has_timer(xprt))
 755                 mod_timer(&xprt->timer, xprt->last_used + xprt->idle_timeout);
 756 }
 757
 758 static void
 759 xprt_init_autodisconnect(struct timer_list *t)
 760 {
 761         struct rpc_xprt *xprt = from_timer(xprt, t, timer);
 762
 763         if (!RB_EMPTY_ROOT(&xprt->recv_queue))
 764                 return;
 765         /* Reset xprt->last_used to avoid connect/autodisconnect cycling */
 766         xprt->last_used = jiffies;
 767         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 768                 return;
 769         queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 770 }
 771
 772 bool xprt_lock_connect(struct rpc_xprt *xprt,
 773                 struct rpc_task *task,
 774                 void *cookie)
 775 {
 776         bool ret = false;
 777
 778         spin_lock(&xprt->transport_lock);
 779         if (!test_bit(XPRT_LOCKED, &xprt->state))
 780                 goto out;
 781         if (xprt->snd_task != task)
 782                 goto out;
 783         xprt->snd_task = cookie;
 784         ret = true;
 785 out:
 786         spin_unlock(&xprt->transport_lock);
 787         return ret;
 788 }
 789
 790 void xprt_unlock_connect(struct rpc_xprt *xprt, void *cookie)
 791 {
 792         spin_lock(&xprt->transport_lock);
 793         if (xprt->snd_task != cookie)
 794                 goto out;
 795         if (!test_bit(XPRT_LOCKED, &xprt->state))
 796                 goto out;
 797         xprt->snd_task =NULL;
 798         xprt->ops->release_xprt(xprt, NULL);
 799         xprt_schedule_autodisconnect(xprt);
 800 out:
 801         spin_unlock(&xprt->transport_lock);
 802         wake_up_bit(&xprt->state, XPRT_LOCKED);
 803 }
 804
 805 /**
 806  * xprt_connect - schedule a transport connect operation
 807  * @task: RPC task that is requesting the connect
 808  *
 809  */
 810 void xprt_connect(struct rpc_task *task)
 811 {
 812         struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 813
 814         dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
 815                         xprt, (xprt_connected(xprt) ? "is" : "is not"));
 816
 817         if (!xprt_bound(xprt)) {
 818                 task->tk_status = -EAGAIN;
 819                 return;
 820         }
 821         if (!xprt_lock_write(xprt, task))
 822                 return;
 823
 824         if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state))
 825                 xprt->ops->close(xprt);
 826
 827         if (!xprt_connected(xprt)) {
 828                 task->tk_rqstp->rq_connect_cookie = xprt->connect_cookie;
 829                 rpc_sleep_on_timeout(&xprt->pending, task, NULL,
 830                                 xprt_request_timeout(task->tk_rqstp));
 831
 832                 if (test_bit(XPRT_CLOSING, &xprt->state))
 833                         return;
 834                 if (xprt_test_and_set_connecting(xprt))
 835                         return;
 836                 /* Race breaker */
 837                 if (!xprt_connected(xprt)) {
 838                         xprt->stat.connect_start = jiffies;
 839                         xprt->ops->connect(xprt, task);
 840                 } else {
 841                         xprt_clear_connecting(xprt);
 842                         task->tk_status = 0;
 843                         rpc_wake_up_queued_task(&xprt->pending, task);
 844                 }
 845         }
 846         xprt_release_write(xprt, task);
 847 }
 848
 849 /**
 850  * xprt_reconnect_delay - compute the wait before scheduling a connect
 851  * @xprt: transport instance
 852  *
 853  */
 854 unsigned long xprt_reconnect_delay(const struct rpc_xprt *xprt)
 855 {
 856         unsigned long start, now = jiffies;
 857
 858         start = xprt->stat.connect_start + xprt->reestablish_timeout;
 859         if (time_after(start, now))
 860                 return start - now;
 861         return 0;
 862 }
 863 EXPORT_SYMBOL_GPL(xprt_reconnect_delay);
 864
 865 /**
 866  * xprt_reconnect_backoff - compute the new re-establish timeout
 867  * @xprt: transport instance
 868  * @init_to: initial reestablish timeout
 869  *
 870  */
 871 void xprt_reconnect_backoff(struct rpc_xprt *xprt, unsigned long init_to)
 872 {
 873         xprt->reestablish_timeout <<= 1;
 874         if (xprt->reestablish_timeout > xprt->max_reconnect_timeout)
 875                 xprt->reestablish_timeout = xprt->max_reconnect_timeout;
 876         if (xprt->reestablish_timeout < init_to)
 877                 xprt->reestablish_timeout = init_to;
 878 }
 879 EXPORT_SYMBOL_GPL(xprt_reconnect_backoff);
 880
 881 enum xprt_xid_rb_cmp {
 882         XID_RB_EQUAL,
 883         XID_RB_LEFT,
 884         XID_RB_RIGHT,
 885 };
 886 static enum xprt_xid_rb_cmp
 887 xprt_xid_cmp(__be32 xid1, __be32 xid2)
 888 {
 889         if (xid1 == xid2)
 890                 return XID_RB_EQUAL;
 891         if ((__force u32)xid1 < (__force u32)xid2)
 892                 return XID_RB_LEFT;
 893         return XID_RB_RIGHT;
 894 }
 895
 896 static struct rpc_rqst *
 897 xprt_request_rb_find(struct rpc_xprt *xprt, __be32 xid)
 898 {
 899         struct rb_node *n = xprt->recv_queue.rb_node;
 900         struct rpc_rqst *req;
 901
 902         while (n != NULL) {
 903                 req = rb_entry(n, struct rpc_rqst, rq_recv);
 904                 switch (xprt_xid_cmp(xid, req->rq_xid)) {
 905                 case XID_RB_LEFT:
 906                         n = n->rb_left;
 907                         break;
 908                 case XID_RB_RIGHT:
 909                         n = n->rb_right;
 910                         break;
 911                 case XID_RB_EQUAL:
 912                         return req;
 913                 }
 914         }
 915         return NULL;
 916 }
 917
 918 static void
 919 xprt_request_rb_insert(struct rpc_xprt *xprt, struct rpc_rqst *new)
 920 {
 921         struct rb_node **p = &xprt->recv_queue.rb_node;
 922         struct rb_node *n = NULL;
 923         struct rpc_rqst *req;
 924
 925         while (*p != NULL) {
 926                 n = *p;
 927                 req = rb_entry(n, struct rpc_rqst, rq_recv);
 928                 switch(xprt_xid_cmp(new->rq_xid, req->rq_xid)) {
 929                 case XID_RB_LEFT:
 930                         p = &n->rb_left;
 931                         break;
 932                 case XID_RB_RIGHT:
 933                         p = &n->rb_right;
 934                         break;
 935                 case XID_RB_EQUAL:
 936                         WARN_ON_ONCE(new != req);
 937                         return;
 938                 }
 939         }
 940         rb_link_node(&new->rq_recv, n, p);
 941         rb_insert_color(&new->rq_recv, &xprt->recv_queue);
 942 }
 943
 944 static void
 945 xprt_request_rb_remove(struct rpc_xprt *xprt, struct rpc_rqst *req)
 946 {
 947         rb_erase(&req->rq_recv, &xprt->recv_queue);
 948 }
 949
 950 /**
 951  * xprt_lookup_rqst - find an RPC request corresponding to an XID
 952  * @xprt: transport on which the original request was transmitted
 953  * @xid: RPC XID of incoming reply
 954  *
 955  * Caller holds xprt->queue_lock.
 956  */
 957 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
 958 {
 959         struct rpc_rqst *entry;
 960
 961         entry = xprt_request_rb_find(xprt, xid);
 962         if (entry != NULL) {
 963                 trace_xprt_lookup_rqst(xprt, xid, 0);
 964                 entry->rq_rtt = ktime_sub(ktime_get(), entry->rq_xtime);
 965                 return entry;
 966         }
 967
 968         dprintk("RPC:       xprt_lookup_rqst did not find xid %08x\n",
 969                         ntohl(xid));
 970         trace_xprt_lookup_rqst(xprt, xid, -ENOENT);
 971         xprt->stat.bad_xids++;
 972         return NULL;
 973 }
 974 EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
 975
 976 static bool
 977 xprt_is_pinned_rqst(struct rpc_rqst *req)
 978 {
 979         return atomic_read(&req->rq_pin) != 0;
 980 }
 981
 982 /**
 983  * xprt_pin_rqst - Pin a request on the transport receive list
 984  * @req: Request to pin
 985  *
 986  * Caller must ensure this is atomic with the call to xprt_lookup_rqst()
 987  * so should be holding xprt->queue_lock.
 988  */
 989 void xprt_pin_rqst(struct rpc_rqst *req)
 990 {
 991         atomic_inc(&req->rq_pin);
 992 }
 993 EXPORT_SYMBOL_GPL(xprt_pin_rqst);
 994
 995 /**
 996  * xprt_unpin_rqst - Unpin a request on the transport receive list
 997  * @req: Request to pin
 998  *
 999  * Caller should be holding xprt->queue_lock.
1000  */
1001 void xprt_unpin_rqst(struct rpc_rqst *req)
1002 {
1003         if (!test_bit(RPC_TASK_MSG_PIN_WAIT, &req->rq_task->tk_runstate)) {
1004                 atomic_dec(&req->rq_pin);
1005                 return;
1006         }
1007         if (atomic_dec_and_test(&req->rq_pin))
1008                 wake_up_var(&req->rq_pin);
1009 }
1010 EXPORT_SYMBOL_GPL(xprt_unpin_rqst);
1011
1012 static void xprt_wait_on_pinned_rqst(struct rpc_rqst *req)
1013 {
1014         wait_var_event(&req->rq_pin, !xprt_is_pinned_rqst(req));
1015 }
1016
1017 static bool
1018 xprt_request_data_received(struct rpc_task *task)
1019 {
1020         return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1021                 READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) != 0;
1022 }
1023
1024 static bool
1025 xprt_request_need_enqueue_receive(struct rpc_task *task, struct rpc_rqst *req)
1026 {
1027         return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1028                 READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) == 0;
1029 }
1030
1031 /**
1032  * xprt_request_enqueue_receive - Add an request to the receive queue
1033  * @task: RPC task
1034  *
1035  */
1036 void
1037 xprt_request_enqueue_receive(struct rpc_task *task)
1038 {
1039         struct rpc_rqst *req = task->tk_rqstp;
1040         struct rpc_xprt *xprt = req->rq_xprt;
1041
1042         if (!xprt_request_need_enqueue_receive(task, req))
1043                 return;
1044
1045         xprt_request_prepare(task->tk_rqstp);
1046         spin_lock(&xprt->queue_lock);
1047
1048         /* Update the softirq receive buffer */
1049         memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
1050                         sizeof(req->rq_private_buf));
1051
1052         /* Add request to the receive list */
1053         xprt_request_rb_insert(xprt, req);
1054         set_bit(RPC_TASK_NEED_RECV, &task->tk_runstate);
1055         spin_unlock(&xprt->queue_lock);
1056
1057         /* Turn off autodisconnect */
1058         del_singleshot_timer_sync(&xprt->timer);
1059 }
1060
1061 /**
1062  * xprt_request_dequeue_receive_locked - Remove a request from the receive queue
1063  * @task: RPC task
1064  *
1065  * Caller must hold xprt->queue_lock.
1066  */
1067 static void
1068 xprt_request_dequeue_receive_locked(struct rpc_task *task)
1069 {
1070         struct rpc_rqst *req = task->tk_rqstp;
1071
1072         if (test_and_clear_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1073                 xprt_request_rb_remove(req->rq_xprt, req);
1074 }
1075
1076 /**
1077  * xprt_update_rtt - Update RPC RTT statistics
1078  * @task: RPC request that recently completed
1079  *
1080  * Caller holds xprt->queue_lock.
1081  */
1082 void xprt_update_rtt(struct rpc_task *task)
1083 {
1084         struct rpc_rqst *req = task->tk_rqstp;
1085         struct rpc_rtt *rtt = task->tk_client->cl_rtt;
1086         unsigned int timer = task->tk_msg.rpc_proc->p_timer;
1087         long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
1088
1089         if (timer) {
1090                 if (req->rq_ntrans == 1)
1091                         rpc_update_rtt(rtt, timer, m);
1092                 rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
1093         }
1094 }
1095 EXPORT_SYMBOL_GPL(xprt_update_rtt);
1096
1097 /**
1098  * xprt_complete_rqst - called when reply processing is complete
1099  * @task: RPC request that recently completed
1100  * @copied: actual number of bytes received from the transport
1101  *
1102  * Caller holds xprt->queue_lock.
1103  */
1104 void xprt_complete_rqst(struct rpc_task *task, int copied)
1105 {
1106         struct rpc_rqst *req = task->tk_rqstp;
1107         struct rpc_xprt *xprt = req->rq_xprt;
1108
1109         dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
1110                         task->tk_pid, ntohl(req->rq_xid), copied);
1111         trace_xprt_complete_rqst(xprt, req->rq_xid, copied);
1112
1113         xprt->stat.recvs++;
1114
1115         req->rq_private_buf.len = copied;
1116         /* Ensure all writes are done before we update */
1117         /* req->rq_reply_bytes_recvd */
1118         smp_wmb();
1119         req->rq_reply_bytes_recvd = copied;
1120         xprt_request_dequeue_receive_locked(task);
1121         rpc_wake_up_queued_task(&xprt->pending, task);
1122 }
1123 EXPORT_SYMBOL_GPL(xprt_complete_rqst);
1124
1125 static void xprt_timer(struct rpc_task *task)
1126 {
1127         struct rpc_rqst *req = task->tk_rqstp;
1128         struct rpc_xprt *xprt = req->rq_xprt;
1129
1130         if (task->tk_status != -ETIMEDOUT)
1131                 return;
1132
1133         trace_xprt_timer(xprt, req->rq_xid, task->tk_status);
1134         if (!req->rq_reply_bytes_recvd) {
1135                 if (xprt->ops->timer)
1136                         xprt->ops->timer(xprt, task);
1137         } else
1138                 task->tk_status = 0;
1139 }
1140
1141 /**
1142  * xprt_wait_for_reply_request_def - wait for reply
1143  * @task: pointer to rpc_task
1144  *
1145  * Set a request's retransmit timeout based on the transport's
1146  * default timeout parameters.  Used by transports that don't adjust
1147  * the retransmit timeout based on round-trip time estimation,
1148  * and put the task to sleep on the pending queue.
1149  */
1150 void xprt_wait_for_reply_request_def(struct rpc_task *task)
1151 {
1152         struct rpc_rqst *req = task->tk_rqstp;
1153
1154         rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1155                         xprt_request_timeout(req));
1156 }
1157 EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_def);
1158
1159 /**
1160  * xprt_wait_for_reply_request_rtt - wait for reply using RTT estimator
1161  * @task: pointer to rpc_task
1162  *
1163  * Set a request's retransmit timeout using the RTT estimator,
1164  * and put the task to sleep on the pending queue.
1165  */
1166 void xprt_wait_for_reply_request_rtt(struct rpc_task *task)
1167 {
1168         int timer = task->tk_msg.rpc_proc->p_timer;
1169         struct rpc_clnt *clnt = task->tk_client;
1170         struct rpc_rtt *rtt = clnt->cl_rtt;
1171         struct rpc_rqst *req = task->tk_rqstp;
1172         unsigned long max_timeout = clnt->cl_timeout->to_maxval;
1173         unsigned long timeout;
1174
1175         timeout = rpc_calc_rto(rtt, timer);
1176         timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
1177         if (timeout > max_timeout || timeout == 0)
1178                 timeout = max_timeout;
1179         rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1180                         jiffies + timeout);
1181 }
1182 EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_rtt);
1183
1184 /**
1185  * xprt_request_wait_receive - wait for the reply to an RPC request
1186  * @task: RPC task about to send a request
1187  *
1188  */
1189 void xprt_request_wait_receive(struct rpc_task *task)
1190 {
1191         struct rpc_rqst *req = task->tk_rqstp;
1192         struct rpc_xprt *xprt = req->rq_xprt;
1193
1194         if (!test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1195                 return;
1196         /*
1197          * Sleep on the pending queue if we're expecting a reply.
1198          * The spinlock ensures atomicity between the test of
1199          * req->rq_reply_bytes_recvd, and the call to rpc_sleep_on().
1200          */
1201         spin_lock(&xprt->queue_lock);
1202         if (test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate)) {
1203                 xprt->ops->wait_for_reply_request(task);
1204                 /*
1205                  * Send an extra queue wakeup call if the
1206                  * connection was dropped in case the call to
1207                  * rpc_sleep_on() raced.
1208                  */
1209                 if (xprt_request_retransmit_after_disconnect(task))
1210                         rpc_wake_up_queued_task_set_status(&xprt->pending,
1211                                         task, -ENOTCONN);
1212         }
1213         spin_unlock(&xprt->queue_lock);
1214 }
1215
1216 static bool
1217 xprt_request_need_enqueue_transmit(struct rpc_task *task, struct rpc_rqst *req)
1218 {
1219         return !test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1220 }
1221
1222 /**
1223  * xprt_request_enqueue_transmit - queue a task for transmission
1224  * @task: pointer to rpc_task
1225  *
1226  * Add a task to the transmission queue.
1227  */
1228 void
1229 xprt_request_enqueue_transmit(struct rpc_task *task)
1230 {
1231         struct rpc_rqst *pos, *req = task->tk_rqstp;
1232         struct rpc_xprt *xprt = req->rq_xprt;
1233
1234         if (xprt_request_need_enqueue_transmit(task, req)) {
1235                 req->rq_bytes_sent = 0;
1236                 spin_lock(&xprt->queue_lock);
1237                 /*
1238                  * Requests that carry congestion control credits are added
1239                  * to the head of the list to avoid starvation issues.
1240                  */
1241                 if (req->rq_cong) {
1242                         xprt_clear_congestion_window_wait(xprt);
1243                         list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1244                                 if (pos->rq_cong)
1245                                         continue;
1246                                 /* Note: req is added _before_ pos */
1247                                 list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1248                                 INIT_LIST_HEAD(&req->rq_xmit2);
1249                                 trace_xprt_enq_xmit(task, 1);
1250                                 goto out;
1251                         }
1252                 } else if (RPC_IS_SWAPPER(task)) {
1253                         list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1254                                 if (pos->rq_cong || pos->rq_bytes_sent)
1255                                         continue;
1256                                 if (RPC_IS_SWAPPER(pos->rq_task))
1257                                         continue;
1258                                 /* Note: req is added _before_ pos */
1259                                 list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1260                                 INIT_LIST_HEAD(&req->rq_xmit2);
1261                                 trace_xprt_enq_xmit(task, 2);
1262                                 goto out;
1263                         }
1264                 } else if (!req->rq_seqno) {
1265                         list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1266                                 if (pos->rq_task->tk_owner != task->tk_owner)
1267                                         continue;
1268                                 list_add_tail(&req->rq_xmit2, &pos->rq_xmit2);
1269                                 INIT_LIST_HEAD(&req->rq_xmit);
1270                                 trace_xprt_enq_xmit(task, 3);
1271                                 goto out;
1272                         }
1273                 }
1274                 list_add_tail(&req->rq_xmit, &xprt->xmit_queue);
1275                 INIT_LIST_HEAD(&req->rq_xmit2);
1276                 trace_xprt_enq_xmit(task, 4);
1277 out:
1278                 set_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1279                 spin_unlock(&xprt->queue_lock);
1280         }
1281 }
1282
1283 /**
1284  * xprt_request_dequeue_transmit_locked - remove a task from the transmission queue
1285  * @task: pointer to rpc_task
1286  *
1287  * Remove a task from the transmission queue
1288  * Caller must hold xprt->queue_lock
1289  */
1290 static void
1291 xprt_request_dequeue_transmit_locked(struct rpc_task *task)
1292 {
1293         struct rpc_rqst *req = task->tk_rqstp;
1294
1295         if (!test_and_clear_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1296                 return;
1297         if (!list_empty(&req->rq_xmit)) {
1298                 list_del(&req->rq_xmit);
1299                 if (!list_empty(&req->rq_xmit2)) {
1300                         struct rpc_rqst *next = list_first_entry(&req->rq_xmit2,
1301                                         struct rpc_rqst, rq_xmit2);
1302                         list_del(&req->rq_xmit2);
1303                         list_add_tail(&next->rq_xmit, &next->rq_xprt->xmit_queue);
1304                 }
1305         } else
1306                 list_del(&req->rq_xmit2);
1307 }
1308
1309 /**
1310  * xprt_request_dequeue_transmit - remove a task from the transmission queue
1311  * @task: pointer to rpc_task
1312  *
1313  * Remove a task from the transmission queue
1314  */
1315 static void
1316 xprt_request_dequeue_transmit(struct rpc_task *task)
1317 {
1318         struct rpc_rqst *req = task->tk_rqstp;
1319         struct rpc_xprt *xprt = req->rq_xprt;
1320
1321         spin_lock(&xprt->queue_lock);
1322         xprt_request_dequeue_transmit_locked(task);
1323         spin_unlock(&xprt->queue_lock);
1324 }
1325
1326 /**
1327  * xprt_request_prepare - prepare an encoded request for transport
1328  * @req: pointer to rpc_rqst
1329  *
1330  * Calls into the transport layer to do whatever is needed to prepare
1331  * the request for transmission or receive.
1332  */
1333 void
1334 xprt_request_prepare(struct rpc_rqst *req)
1335 {
1336         struct rpc_xprt *xprt = req->rq_xprt;
1337
1338         if (xprt->ops->prepare_request)
1339                 xprt->ops->prepare_request(req);
1340 }
1341
1342 /**
1343  * xprt_request_need_retransmit - Test if a task needs retransmission
1344  * @task: pointer to rpc_task
1345  *
1346  * Test for whether a connection breakage requires the task to retransmit
1347  */
1348 bool
1349 xprt_request_need_retransmit(struct rpc_task *task)
1350 {
1351         return xprt_request_retransmit_after_disconnect(task);
1352 }
1353
1354 /**
1355  * xprt_prepare_transmit - reserve the transport before sending a request
1356  * @task: RPC task about to send a request
1357  *
1358  */
1359 bool xprt_prepare_transmit(struct rpc_task *task)
1360 {
1361         struct rpc_rqst *req = task->tk_rqstp;
1362         struct rpc_xprt *xprt = req->rq_xprt;
1363
1364         dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
1365
1366         if (!xprt_lock_write(xprt, task)) {
1367                 /* Race breaker: someone may have transmitted us */
1368                 if (!test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1369                         rpc_wake_up_queued_task_set_status(&xprt->sending,
1370                                         task, 0);
1371                 return false;
1372
1373         }
1374         return true;
1375 }
1376
1377 void xprt_end_transmit(struct rpc_task *task)
1378 {
1379         xprt_release_write(task->tk_rqstp->rq_xprt, task);
1380 }
1381
1382 /**
1383  * xprt_request_transmit - send an RPC request on a transport
1384  * @req: pointer to request to transmit
1385  * @snd_task: RPC task that owns the transport lock
1386  *
1387  * This performs the transmission of a single request.
1388  * Note that if the request is not the same as snd_task, then it
1389  * does need to be pinned.
1390  * Returns '0' on success.
1391  */
1392 static int
1393 xprt_request_transmit(struct rpc_rqst *req, struct rpc_task *snd_task)
1394 {
1395         struct rpc_xprt *xprt = req->rq_xprt;
1396         struct rpc_task *task = req->rq_task;
1397         unsigned int connect_cookie;
1398         int is_retrans = RPC_WAS_SENT(task);
1399         int status;
1400
1401         if (!req->rq_bytes_sent) {
1402                 if (xprt_request_data_received(task)) {
1403                         status = 0;
1404                         goto out_dequeue;
1405                 }
1406                 /* Verify that our message lies in the RPCSEC_GSS window */
1407                 if (rpcauth_xmit_need_reencode(task)) {
1408                         status = -EBADMSG;
1409                         goto out_dequeue;
1410                 }
1411                 if (RPC_SIGNALLED(task)) {
1412                         status = -ERESTARTSYS;
1413                         goto out_dequeue;
1414                 }
1415         }
1416
1417         /*
1418          * Update req->rq_ntrans before transmitting to avoid races with
1419          * xprt_update_rtt(), which needs to know that it is recording a
1420          * reply to the first transmission.
1421          */
1422         req->rq_ntrans++;
1423
1424         connect_cookie = xprt->connect_cookie;
1425         status = xprt->ops->send_request(req);
1426         if (status != 0) {
1427                 req->rq_ntrans--;
1428                 trace_xprt_transmit(req, status);
1429                 return status;
1430         }
1431
1432         if (is_retrans)
1433                 task->tk_client->cl_stats->rpcretrans++;
1434
1435         xprt_inject_disconnect(xprt);
1436
1437         task->tk_flags |= RPC_TASK_SENT;
1438         spin_lock(&xprt->transport_lock);
1439
1440         xprt->stat.sends++;
1441         xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
1442         xprt->stat.bklog_u += xprt->backlog.qlen;
1443         xprt->stat.sending_u += xprt->sending.qlen;
1444         xprt->stat.pending_u += xprt->pending.qlen;
1445         spin_unlock(&xprt->transport_lock);
1446
1447         req->rq_connect_cookie = connect_cookie;
1448 out_dequeue:
1449         trace_xprt_transmit(req, status);
1450         xprt_request_dequeue_transmit(task);
1451         rpc_wake_up_queued_task_set_status(&xprt->sending, task, status);
1452         return status;
1453 }
1454
1455 /**
1456  * xprt_transmit - send an RPC request on a transport
1457  * @task: controlling RPC task
1458  *
1459  * Attempts to drain the transmit queue. On exit, either the transport
1460  * signalled an error that needs to be handled before transmission can
1461  * resume, or @task finished transmitting, and detected that it already
1462  * received a reply.
1463  */
1464 void
1465 xprt_transmit(struct rpc_task *task)
1466 {
1467         struct rpc_rqst *next, *req = task->tk_rqstp;
1468         struct rpc_xprt *xprt = req->rq_xprt;
1469         int status;
1470
1471         spin_lock(&xprt->queue_lock);
1472         while (!list_empty(&xprt->xmit_queue)) {
1473                 next = list_first_entry(&xprt->xmit_queue,
1474                                 struct rpc_rqst, rq_xmit);
1475                 xprt_pin_rqst(next);
1476                 spin_unlock(&xprt->queue_lock);
1477                 status = xprt_request_transmit(next, task);
1478                 if (status == -EBADMSG && next != req)
1479                         status = 0;
1480                 cond_resched();
1481                 spin_lock(&xprt->queue_lock);
1482                 xprt_unpin_rqst(next);
1483                 if (status == 0) {
1484                         if (!xprt_request_data_received(task) ||
1485                             test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1486                                 continue;
1487                 } else if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1488                         task->tk_status = status;
1489                 break;
1490         }
1491         spin_unlock(&xprt->queue_lock);
1492 }
1493
1494 static void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
1495 {
1496         set_bit(XPRT_CONGESTED, &xprt->state);
1497         rpc_sleep_on(&xprt->backlog, task, NULL);
1498 }
1499
1500 static void xprt_wake_up_backlog(struct rpc_xprt *xprt)
1501 {
1502         if (rpc_wake_up_next(&xprt->backlog) == NULL)
1503                 clear_bit(XPRT_CONGESTED, &xprt->state);
1504 }
1505
1506 static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task)
1507 {
1508         bool ret = false;
1509
1510         if (!test_bit(XPRT_CONGESTED, &xprt->state))
1511                 goto out;
1512         spin_lock(&xprt->reserve_lock);
1513         if (test_bit(XPRT_CONGESTED, &xprt->state)) {
1514                 rpc_sleep_on(&xprt->backlog, task, NULL);
1515                 ret = true;
1516         }
1517         spin_unlock(&xprt->reserve_lock);
1518 out:
1519         return ret;
1520 }
1521
1522 static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt)
1523 {
1524         struct rpc_rqst *req = ERR_PTR(-EAGAIN);
1525
1526         if (xprt->num_reqs >= xprt->max_reqs)
1527                 goto out;
1528         ++xprt->num_reqs;
1529         spin_unlock(&xprt->reserve_lock);
1530         req = kzalloc(sizeof(struct rpc_rqst), GFP_NOFS);
1531         spin_lock(&xprt->reserve_lock);
1532         if (req != NULL)
1533                 goto out;
1534         --xprt->num_reqs;
1535         req = ERR_PTR(-ENOMEM);
1536 out:
1537         return req;
1538 }
1539
1540 static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1541 {
1542         if (xprt->num_reqs > xprt->min_reqs) {
1543                 --xprt->num_reqs;
1544                 kfree(req);
1545                 return true;
1546         }
1547         return false;
1548 }
1549
1550 void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
1551 {
1552         struct rpc_rqst *req;
1553
1554         spin_lock(&xprt->reserve_lock);
1555         if (!list_empty(&xprt->free)) {
1556                 req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
1557                 list_del(&req->rq_list);
1558                 goto out_init_req;
1559         }
1560         req = xprt_dynamic_alloc_slot(xprt);
1561         if (!IS_ERR(req))
1562                 goto out_init_req;
1563         switch (PTR_ERR(req)) {
1564         case -ENOMEM:
1565                 dprintk("RPC:       dynamic allocation of request slot "
1566                                 "failed! Retrying\n");
1567                 task->tk_status = -ENOMEM;
1568                 break;
1569         case -EAGAIN:
1570                 xprt_add_backlog(xprt, task);
1571                 dprintk("RPC:       waiting for request slot\n");
1572                 /* fall through */
1573         default:
1574                 task->tk_status = -EAGAIN;
1575         }
1576         spin_unlock(&xprt->reserve_lock);
1577         return;
1578 out_init_req:
1579         xprt->stat.max_slots = max_t(unsigned int, xprt->stat.max_slots,
1580                                      xprt->num_reqs);
1581         spin_unlock(&xprt->reserve_lock);
1582
1583         task->tk_status = 0;
1584         task->tk_rqstp = req;
1585 }
1586 EXPORT_SYMBOL_GPL(xprt_alloc_slot);
1587
1588 void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1589 {
1590         spin_lock(&xprt->reserve_lock);
1591         if (!xprt_dynamic_free_slot(xprt, req)) {
1592                 memset(req, 0, sizeof(*req));   /* mark unused */
1593                 list_add(&req->rq_list, &xprt->free);
1594         }
1595         xprt_wake_up_backlog(xprt);
1596         spin_unlock(&xprt->reserve_lock);
1597 }
1598 EXPORT_SYMBOL_GPL(xprt_free_slot);
1599
1600 static void xprt_free_all_slots(struct rpc_xprt *xprt)
1601 {
1602         struct rpc_rqst *req;
1603         while (!list_empty(&xprt->free)) {
1604                 req = list_first_entry(&xprt->free, struct rpc_rqst, rq_list);
1605                 list_del(&req->rq_list);
1606                 kfree(req);
1607         }
1608 }
1609
1610 struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
1611                 unsigned int num_prealloc,
1612                 unsigned int max_alloc)
1613 {
1614         struct rpc_xprt *xprt;
1615         struct rpc_rqst *req;
1616         int i;
1617
1618         xprt = kzalloc(size, GFP_KERNEL);
1619         if (xprt == NULL)
1620                 goto out;
1621
1622         xprt_init(xprt, net);
1623
1624         for (i = 0; i < num_prealloc; i++) {
1625                 req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
1626                 if (!req)
1627                         goto out_free;
1628                 list_add(&req->rq_list, &xprt->free);
1629         }
1630         if (max_alloc > num_prealloc)
1631                 xprt->max_reqs = max_alloc;
1632         else
1633                 xprt->max_reqs = num_prealloc;
1634         xprt->min_reqs = num_prealloc;
1635         xprt->num_reqs = num_prealloc;
1636
1637         return xprt;
1638
1639 out_free:
1640         xprt_free(xprt);
1641 out:
1642         return NULL;
1643 }
1644 EXPORT_SYMBOL_GPL(xprt_alloc);
1645
1646 void xprt_free(struct rpc_xprt *xprt)
1647 {
1648         put_net(xprt->xprt_net);
1649         xprt_free_all_slots(xprt);
1650         kfree_rcu(xprt, rcu);
1651 }
1652 EXPORT_SYMBOL_GPL(xprt_free);
1653
1654 static void
1655 xprt_init_connect_cookie(struct rpc_rqst *req, struct rpc_xprt *xprt)
1656 {
1657         req->rq_connect_cookie = xprt_connect_cookie(xprt) - 1;
1658 }
1659
1660 static __be32
1661 xprt_alloc_xid(struct rpc_xprt *xprt)
1662 {
1663         __be32 xid;
1664
1665         spin_lock(&xprt->reserve_lock);
1666         xid = (__force __be32)xprt->xid++;
1667         spin_unlock(&xprt->reserve_lock);
1668         return xid;
1669 }
1670
1671 static void
1672 xprt_init_xid(struct rpc_xprt *xprt)
1673 {
1674         xprt->xid = prandom_u32();
1675 }
1676
1677 static void
1678 xprt_request_init(struct rpc_task *task)
1679 {
1680         struct rpc_xprt *xprt = task->tk_xprt;
1681         struct rpc_rqst *req = task->tk_rqstp;
1682
1683         req->rq_task    = task;
1684         req->rq_xprt    = xprt;
1685         req->rq_buffer  = NULL;
1686         req->rq_xid     = xprt_alloc_xid(xprt);
1687         xprt_init_connect_cookie(req, xprt);
1688         req->rq_snd_buf.len = 0;
1689         req->rq_snd_buf.buflen = 0;
1690         req->rq_rcv_buf.len = 0;
1691         req->rq_rcv_buf.buflen = 0;
1692         req->rq_snd_buf.bvec = NULL;
1693         req->rq_rcv_buf.bvec = NULL;
1694         req->rq_release_snd_buf = NULL;
1695         xprt_init_majortimeo(task, req);
1696         dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
1697                         req, ntohl(req->rq_xid));
1698 }
1699
1700 static void
1701 xprt_do_reserve(struct rpc_xprt *xprt, struct rpc_task *task)
1702 {
1703         xprt->ops->alloc_slot(xprt, task);
1704         if (task->tk_rqstp != NULL)
1705                 xprt_request_init(task);
1706 }
1707
1708 /**
1709  * xprt_reserve - allocate an RPC request slot
1710  * @task: RPC task requesting a slot allocation
1711  *
1712  * If the transport is marked as being congested, or if no more
1713  * slots are available, place the task on the transport's
1714  * backlog queue.
1715  */
1716 void xprt_reserve(struct rpc_task *task)
1717 {
1718         struct rpc_xprt *xprt = task->tk_xprt;
1719
1720         task->tk_status = 0;
1721         if (task->tk_rqstp != NULL)
1722                 return;
1723
1724         task->tk_status = -EAGAIN;
1725         if (!xprt_throttle_congested(xprt, task))
1726                 xprt_do_reserve(xprt, task);
1727 }
1728
1729 /**
1730  * xprt_retry_reserve - allocate an RPC request slot
1731  * @task: RPC task requesting a slot allocation
1732  *
1733  * If no more slots are available, place the task on the transport's
1734  * backlog queue.
1735  * Note that the only difference with xprt_reserve is that we now
1736  * ignore the value of the XPRT_CONGESTED flag.
1737  */
1738 void xprt_retry_reserve(struct rpc_task *task)
1739 {
1740         struct rpc_xprt *xprt = task->tk_xprt;
1741
1742         task->tk_status = 0;
1743         if (task->tk_rqstp != NULL)
1744                 return;
1745
1746         task->tk_status = -EAGAIN;
1747         xprt_do_reserve(xprt, task);
1748 }
1749
1750 static void
1751 xprt_request_dequeue_all(struct rpc_task *task, struct rpc_rqst *req)
1752 {
1753         struct rpc_xprt *xprt = req->rq_xprt;
1754
1755         if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate) ||
1756             test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) ||
1757             xprt_is_pinned_rqst(req)) {
1758                 spin_lock(&xprt->queue_lock);
1759                 xprt_request_dequeue_transmit_locked(task);
1760                 xprt_request_dequeue_receive_locked(task);
1761                 while (xprt_is_pinned_rqst(req)) {
1762                         set_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1763                         spin_unlock(&xprt->queue_lock);
1764                         xprt_wait_on_pinned_rqst(req);
1765                         spin_lock(&xprt->queue_lock);
1766                         clear_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1767                 }
1768                 spin_unlock(&xprt->queue_lock);
1769         }
1770 }
1771
1772 /**
1773  * xprt_release - release an RPC request slot
1774  * @task: task which is finished with the slot
1775  *
1776  */
1777 void xprt_release(struct rpc_task *task)
1778 {
1779         struct rpc_xprt *xprt;
1780         struct rpc_rqst *req = task->tk_rqstp;
1781
1782         if (req == NULL) {
1783                 if (task->tk_client) {
1784                         xprt = task->tk_xprt;
1785                         xprt_release_write(xprt, task);
1786                 }
1787                 return;
1788         }
1789
1790         xprt = req->rq_xprt;
1791         xprt_request_dequeue_all(task, req);
1792         spin_lock(&xprt->transport_lock);
1793         xprt->ops->release_xprt(xprt, task);
1794         if (xprt->ops->release_request)
1795                 xprt->ops->release_request(task);
1796         xprt_schedule_autodisconnect(xprt);
1797         spin_unlock(&xprt->transport_lock);
1798         if (req->rq_buffer)
1799                 xprt->ops->buf_free(task);
1800         xprt_inject_disconnect(xprt);
1801         xdr_free_bvec(&req->rq_rcv_buf);
1802         xdr_free_bvec(&req->rq_snd_buf);
1803         if (req->rq_cred != NULL)
1804                 put_rpccred(req->rq_cred);
1805         task->tk_rqstp = NULL;
1806         if (req->rq_release_snd_buf)
1807                 req->rq_release_snd_buf(req);
1808
1809         dprintk("RPC: %5u release request %p\n", task->tk_pid, req);
1810         if (likely(!bc_prealloc(req)))
1811                 xprt->ops->free_slot(xprt, req);
1812         else
1813                 xprt_free_bc_request(req);
1814 }
1815
1816 #ifdef CONFIG_SUNRPC_BACKCHANNEL
1817 void
1818 xprt_init_bc_request(struct rpc_rqst *req, struct rpc_task *task)
1819 {
1820         struct xdr_buf *xbufp = &req->rq_snd_buf;
1821
1822         task->tk_rqstp = req;
1823         req->rq_task = task;
1824         xprt_init_connect_cookie(req, req->rq_xprt);
1825         /*
1826          * Set up the xdr_buf length.
1827          * This also indicates that the buffer is XDR encoded already.
1828          */
1829         xbufp->len = xbufp->head[0].iov_len + xbufp->page_len +
1830                 xbufp->tail[0].iov_len;
1831 }
1832 #endif
1833
1834 static void xprt_init(struct rpc_xprt *xprt, struct net *net)
1835 {
1836         kref_init(&xprt->kref);
1837
1838         spin_lock_init(&xprt->transport_lock);
1839         spin_lock_init(&xprt->reserve_lock);
1840         spin_lock_init(&xprt->queue_lock);
1841
1842         INIT_LIST_HEAD(&xprt->free);
1843         xprt->recv_queue = RB_ROOT;
1844         INIT_LIST_HEAD(&xprt->xmit_queue);
1845 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1846         spin_lock_init(&xprt->bc_pa_lock);
1847         INIT_LIST_HEAD(&xprt->bc_pa_list);
1848 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1849         INIT_LIST_HEAD(&xprt->xprt_switch);
1850
1851         xprt->last_used = jiffies;
1852         xprt->cwnd = RPC_INITCWND;
1853         xprt->bind_index = 0;
1854
1855         rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1856         rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1857         rpc_init_wait_queue(&xprt->sending, "xprt_sending");
1858         rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1859
1860         xprt_init_xid(xprt);
1861
1862         xprt->xprt_net = get_net(net);
1863 }
1864
1865 /**
1866  * xprt_create_transport - create an RPC transport
1867  * @args: rpc transport creation arguments
1868  *
1869  */
1870 struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1871 {
1872         struct rpc_xprt *xprt;
1873         struct xprt_class *t;
1874
1875         spin_lock(&xprt_list_lock);
1876         list_for_each_entry(t, &xprt_list, list) {
1877                 if (t->ident == args->ident) {
1878                         spin_unlock(&xprt_list_lock);
1879                         goto found;
1880                 }
1881         }
1882         spin_unlock(&xprt_list_lock);
1883         dprintk("RPC: transport (%d) not supported\n", args->ident);
1884         return ERR_PTR(-EIO);
1885
1886 found:
1887         xprt = t->setup(args);
1888         if (IS_ERR(xprt)) {
1889                 dprintk("RPC:       xprt_create_transport: failed, %ld\n",
1890                                 -PTR_ERR(xprt));
1891                 goto out;
1892         }
1893         if (args->flags & XPRT_CREATE_NO_IDLE_TIMEOUT)
1894                 xprt->idle_timeout = 0;
1895         INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1896         if (xprt_has_timer(xprt))
1897                 timer_setup(&xprt->timer, xprt_init_autodisconnect, 0);
1898         else
1899                 timer_setup(&xprt->timer, NULL, 0);
1900
1901         if (strlen(args->servername) > RPC_MAXNETNAMELEN) {
1902                 xprt_destroy(xprt);
1903                 return ERR_PTR(-EINVAL);
1904         }
1905         xprt->servername = kstrdup(args->servername, GFP_KERNEL);
1906         if (xprt->servername == NULL) {
1907                 xprt_destroy(xprt);
1908                 return ERR_PTR(-ENOMEM);
1909         }
1910
1911         rpc_xprt_debugfs_register(xprt);
1912
1913         dprintk("RPC:       created transport %p with %u slots\n", xprt,
1914                         xprt->max_reqs);
1915 out:
1916         return xprt;
1917 }
1918
1919 static void xprt_destroy_cb(struct work_struct *work)
1920 {
1921         struct rpc_xprt *xprt =
1922                 container_of(work, struct rpc_xprt, task_cleanup);
1923
1924         rpc_xprt_debugfs_unregister(xprt);
1925         rpc_destroy_wait_queue(&xprt->binding);
1926         rpc_destroy_wait_queue(&xprt->pending);
1927         rpc_destroy_wait_queue(&xprt->sending);
1928         rpc_destroy_wait_queue(&xprt->backlog);
1929         kfree(xprt->servername);
1930         /*
1931          * Tear down transport state and free the rpc_xprt
1932          */
1933         xprt->ops->destroy(xprt);
1934 }
1935
1936 /**
1937  * xprt_destroy - destroy an RPC transport, killing off all requests.
1938  * @xprt: transport to destroy
1939  *
1940  */
1941 static void xprt_destroy(struct rpc_xprt *xprt)
1942 {
1943         dprintk("RPC:       destroying transport %p\n", xprt);
1944
1945         /*
1946          * Exclude transport connect/disconnect handlers and autoclose
1947          */
1948         wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_UNINTERRUPTIBLE);
1949
1950         del_timer_sync(&xprt->timer);
1951
1952         /*
1953          * Destroy sockets etc from the system workqueue so they can
1954          * safely flush receive work running on rpciod.
1955          */
1956         INIT_WORK(&xprt->task_cleanup, xprt_destroy_cb);
1957         schedule_work(&xprt->task_cleanup);
1958 }
1959
1960 static void xprt_destroy_kref(struct kref *kref)
1961 {
1962         xprt_destroy(container_of(kref, struct rpc_xprt, kref));
1963 }
1964
1965 /**
1966  * xprt_get - return a reference to an RPC transport.
1967  * @xprt: pointer to the transport
1968  *
1969  */
1970 struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1971 {
1972         if (xprt != NULL && kref_get_unless_zero(&xprt->kref))
1973                 return xprt;
1974         return NULL;
1975 }
1976 EXPORT_SYMBOL_GPL(xprt_get);
1977
1978 /**
1979  * xprt_put - release a reference to an RPC transport.
1980  * @xprt: pointer to the transport
1981  *
1982  */
1983 void xprt_put(struct rpc_xprt *xprt)
1984 {
1985         if (xprt != NULL)
1986                 kref_put(&xprt->kref, xprt_destroy_kref);
1987 }
1988 EXPORT_SYMBOL_GPL(xprt_put);